a trading strategy based on Elliott Wave Theory - page 226
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
Also, for H-volatility you should have used not a but b,
then it would really be H-volatility, otherwise it is something else again.
I used a tick chart for analysis. And you must have used a minute chart. I think that's why there is a discrepancy in the obtained results. On the second point, I beg to differ. In the dissertation H-volatility is defined as the ratio of the sum of all absolute price movements to ... I.e. in sense these increments are first differences and this is a and not b.
to Yurixx
Yura, these are the patterns... Have you read the relevant place in Pastukhov's thesis? Maybe, going sequentially, one should deal with Markov chains first, and then go on to more complicated things. Especially since Pastukhov proves that it is in principle possible to get arbitrage income from Markov processes. And he is unlikely to have pre-optimized TS on historical data :-)
And another question: where is your Mathcad? As I understand it, there is no problem with the distribution, in fact in the nearest shop, the cracked CD costs about 120 rubles.
The number 2 is the same for many other people who have calculated this parameter.
Also, for H-volatility you should not use a, but b,
then it would really be H-volatility, otherwise it is something else again.
I used a tick chart for my analysis. And you probably used a minute chart. I think the discrepancy in the results is due to that. On the second point, I beg to differ. In the dissertation H-volatility is defined as the ratio of the sum of all absolute price movements to ... i.e. in the sense these increments are the first differences, and this is a, not b.
No, I and Pastukhov and ForAxel (I'm not sure here, but it doesn't matter) and some other people
used ticks. The results are all about the same - about 2. And this is not the first difference,
so you get different results.
In other words, you didn't pre-stabilise the row?
There is no stationarity requirement in the original definition of the method.
At least I haven't found it.
Formally your (b(i)-b(i-1)) - is the difference, but it's the difference between adjacent values
in the original series, and in Pastukhov it is adjacent values multiple of H. Again, formally no one
forbid to have h=0.0001, but in the region of small values of h one usually observes
artifacts of approximately this kind (it is marked with a blue dot in the picture):
I understand that it would be more reasonable to perform H-partitioning of the initial tick series,
for example with h=0.0010. And already to this series to apply FAC and other.
I agree with the last point.
P.S. I see you have corrected the picture!
In the area of "small values" your result tends to mine - 1.35!
And it is not an artifact, but the reality given to us.
This Shepherdian first - "adjacent values multiples of H" and then calculating H-volatility as the sum of the moduli of the increments of multiples of H is, in fact, the residualisation and centring of the synthetic series.
Changed the figure to a more appropriate one.
Residualization - let it be so, but at the level of multiples of H, you have it at the level of pips,
this is the difference. I already wrote, I repeat, in essence H-fractionation is an analogue of the simplest
noise suppression. It's already a different series.
P.S. I see you corrected the picture!
In the area of "small values" your result tends to mine - 1.35!
And it is not an artefact but a reality given to us.
It's not the actual values, it's a conventional diagrammatic representation, it shows conventionally
how this parameter usually behaves depending on the value of H-partitioning.
The figure can be different, and it depends on different things, on different rows of data.
Here are some thoughts on the subject http://forex.kbpauk.ru/showflat.php?Cat=0&Board=mts&Number=139469&page=0&fpart=3
In general, so far, no stable patterns have been detected in the area of small values,
except that they're different from the theoretical ones.
Also, they are not that interesting as they are usually smaller than the spread.
Check it out. I'm quite satisfied.
You may be right - there's no hurry. Well, if I'm going to be consistent, then before dealing with Markov chains I, personally, still have to deal with some very elementary things. For example, what is it? In short, study probability theory and mathematical statistics. I'm afraid it's too elementary, I might not be able to cope !
I've got Mathcad up and running. But it has one drawback - it doesn't do anything without me.
And I can't take part in this process yet. Before you do anything, you first need to understand what, and then how. In the field of statistical research my understanding of it remains at the level of a housewife, so my efforts are directed a bit in a different direction, where I understand more.
If we are speaking about those indicators of BL and BR power about which I wrote earlier, I have come to a conclusion that there is no profit in using them and we must dig in another place. By the way, this conclusion is speculative because I do not possess statistical estimations confirming it. And I have built the distribution of the indicator-price change axis (as you suggested). The result: the distribution of price changes is almost independent of the indicator value, for which the sample is drawn.
It doesn't have to be a Markov process, you can just talk about Markov chains.
You can even talk about chains.